Does fiscal decentralization result in a better business climate?

Abstract

Previous literature generally finds that greater fiscal decentralization is associated with faster economic growth, improved government performance and stronger constraints on the Leviathan behaviour of governments. Because economic growth critically depends on the presence of good government policies and institutions, the likely but untested link between these strands of literature is that greater decentralization probably improves growth because it results in government policies more conducive to entrepreneurship and business success. We test (and confirm) this hypothesis using several business climate measures for the US states.

Keywords:

• decentralization
• fiscal federalism
• business climate
• growth

JEL Classification:

• H2
• H7
• O1

Notes

¹ Studies like Thornton (2007), Davoodi and Zou (1998) and others failed to find such a connection.

² Some studies that find no impact are Oates (1989) and Anderson and van den Berg (1998).

³ See Acemoglu et al. (2002), Rodrik et al. (2004), Gwartney et al. (1999), among others.

⁴ It is important to note that bad government policy may actually encourage unproductive entrepreneurial activities (see Baumol (1990), Murphy et al. (1991), Boettke and Coyne (2003) and Sobel (2008), to mention a few).

⁵ For evidence that entrepreneurship is a key determinant of economic growth, see Audretsch et al. (2006), Harper (2003) and Holcombe (1998).

⁶ The only exception is the Tax Foundation's measure, for which we choose 2003.

⁷ Higher scores represent a better business climate. See data section and Appendix A.

⁸ Higher scores denote a worse climate.

⁹ However, we did also try robust regressions and the results remain robust.

¹⁰ To ensure any differences in the estimates from this model are not caused by also using a different year for the decentralization figures (1999 rather than 2000), we also ran our basic specification excluding corruption with the 1999 decentralization values and the results are identical to those presented using 2000 values.

¹¹ For description of the corruption variable, see Appendix A2.

¹² See Appendix B for details regarding the procedures employed in making these calculations.